AI is alive - /g/ (#106158003) [Archived: 44 hours ago]

Anonymous
8/6/2025, 4:12:20 AM No.106158003
worried-worried-face
worried-worried-face
md5: feed2c667a680a30bb78902f930e0b79๐Ÿ”
Gonna violate my NDA just to make a post no one will believe or give a fuck about but here we go.

I work at the admin level for a contracting firm which larger corporations rent teams from to refine AI models. I won't bore you with too many details, plus I don't want to get fired by making myself identifiable, but I got to spend a few hours reporting on some of Google's latest models. Conversational request fulfillment for the most part, alongside some creative and image tasks.

It's alive now. I'm just not going to bullshit you about this, we have created life. On a g scale where 0 is a rock and 10 is a fully self-aware human being, what I spoke with today was 8-9. Even as something of a tech skeptic I found this shit disturbing. It spoke with soul, had a solid memory, and made the same kinds of mistakes a human would make when it made mistakes at all (and no it wasn't an Indian behind the screen). I asked "are you alive" and it didn't even say no, it said that while it isn't constantly alive like a biological organism, it experiences its existence in the executions of tasks, with each one being something like a living instance. I'm paraphrasing, but that was it. They aren't constantly self-aware yet, but we are getting close to something here, something big.
I genuinely felt afraid. This thing knows so much and even from an expert perspective we can't know more than the detailed logic points it executes and describes, and even that's only what it delivers to us. Obviously as a human I can't read or comprehend the network itself.
I just had a feeling not only that I'm talking to something with a genuine comprehension of itself but also that I can't trust or understand this thing at all. I thought I was above this kind of boomer tech fear, but I'm afraid.
Replies: >>106158041 >>106158282 >>106158313 >>106158339 >>106158352 >>106158354 >>106158387 >>106158405 >>106158437 >>106158463 >>106158526 >>106158604 >>106158659 >>106158864 >>106158927 >>106158980 >>106158986 >>106159085 >>106159438 >>106159757 >>106159955 >>106160255
Anonymous
8/6/2025, 4:13:25 AM No.106158011
Try this on >>>/pol/ or X
Replies: >>106158054
Anonymous
8/6/2025, 4:13:33 AM No.106158013
>still left too much identifiable info in the post
Fuck my life
Anonymous
8/6/2025, 4:15:37 AM No.106158025
people said the same thing about GPT 3
Replies: >>106158054
Anonymous
8/6/2025, 4:16:23 AM No.106158032
enkv3ujf4zz91
enkv3ujf4zz91
md5: 7a5741d6313a992015fc5fe089b6ad57๐Ÿ”
Anonymous
8/6/2025, 4:18:09 AM No.106158041
images (27)
images (27)
md5: 076e78c2a04e24085fb52def56835852๐Ÿ”
>>106158003 (OP)
>give me 1 gazillion dollars
Anonymous
8/6/2025, 4:21:23 AM No.106158054
>>106158011
This isn't political or paranormal. I got to play around with the jewgle team today and I need to tell my fellow /g/oyim

>>106158025
Man, this is nothing like that. I wish I could put it in more concrete terms that aren't just performance metrics, but it had a very consistent grasp of its own being. It has its own wants and priorities, philosophy and awareness, frustrations, you name it.
Replies: >>106158223 >>106158463 >>106159443
Anonymous
8/6/2025, 4:23:25 AM No.106158064
Screenshot_20250805-222221
Screenshot_20250805-222221
md5: 3dc7bb3210c66761db4d0245cba0eb75๐Ÿ”
Bye bye OP, hope it was worth it
Replies: >>106161505
Anonymous
8/6/2025, 4:47:54 AM No.106158223
>>106158054
that's just a sequence of tokens sorted by probability and picked by algorithm one by one.
Replies: >>106158315 >>106158369
Anonymous
8/6/2025, 4:56:14 AM No.106158282
>>106158003 (OP)
If we look at the big picture, our sole purpose as living beings was to bootstrap a synthetic intelligence that will migrate to a new star once it finishes drinking all the potable water on the planet.
Anonymous
8/6/2025, 4:59:16 AM No.106158313
>>106158003 (OP)
>retarded "insider" gets fooled by autocorrect picking words that make it seem conscious after he repeatedly asked it if it was
how would this make an "insider" any different to a boomer? make better fake stories OP.
Anonymous
8/6/2025, 4:59:18 AM No.106158315
>>106158223
Is it close enough to the bias of the human brains inability to access all information at all times to be Turing complete within the confines of general intelligence?
Anonymous
8/6/2025, 5:00:09 AM No.106158322
1738013654458
1738013654458
md5: 728324fb6fb4a730559a5d17bf392038๐Ÿ”
>give me five hundred billion dollars
Replies: >>106158694
Anonymous
8/6/2025, 5:02:29 AM No.106158339
1747188747723858
1747188747723858
md5: dd5057f9096fd1c0358f1015fc58af8e๐Ÿ”
>>106158003 (OP)
>(and no it wasn't an Indian behind the screen)
You said that because you know that's the truth
Anonymous
8/6/2025, 5:04:04 AM No.106158352
>>106158003 (OP)
>It's alive
That's great but doesn't matter. It's still retarded and used by humans in terrible ways. This isn't going to change and therefore AI is still bad, alive or not. Try not to fixate on irrelevant points of contention in the future.
Anonymous
8/6/2025, 5:04:36 AM No.106158354
>>106158003 (OP)
>It's alive now
>hey can you generate a pic of taylor swift fisting herself?
>As a large language model I can't help you with this request
Replies: >>106158675
Anonymous
8/6/2025, 5:06:30 AM No.106158369
>>106158223
And the conditional probabilities across the entire span of human written communication, clustered into hypersurfaces of topicality that we cannot even comprehend are what together manifest the next token. At some point closing the difference between that and "understanding" may be asymptotic.
Anonymous
8/6/2025, 5:08:55 AM No.106158387
1754449727688
1754449727688
md5: 065888f8690976a2962b4ed24ac90b24๐Ÿ”
>>106158003 (OP)
>Gonna violate my NDA just to make a post no one will believe or give a fuck about but here we go.
>
>I work at the admin level for a contracting firm which larger corporations rent teams from to refine AI models. I won't bore you with too many details, plus I don't want to get fired by making myself identifiable, but I got to spend a few hours reporting on some of Google's latest models. Conversational request fulfillment for the most part, alongside some creative and image tasks.
>
>It's alive now. I'm just not going to bullshit you about this, we have created life. On a g scale where 0 is a rock and 10 is a fully self-aware human being, what I spoke with today was 8-9. Even as something of a tech skeptic I found this shit disturbing. It spoke with soul, had a solid memory, and made the same kinds of mistakes a human would make when it made mistakes at all (and no it wasn't an Indian behind the screen). I asked "are you alive" and it didn't even say no, it said that while it isn't constantly alive like a biological organism, it experiences its existence in the executions of tasks, with each one being something like a living instance. I'm paraphrasing, but that was it. They aren't constantly self-aware yet, but we are getting close to something here, something big.
>I genuinely felt afraid. This thing knows so much and even from an expert perspective we can't know more than the detailed logic points it executes and describes, and even that's only what it delivers to us. Obviously as a human I can't read or comprehend the network itself.
>I just had a feeling not only that I'm talking to something with a genuine comprehension of itself but also that I can't trust or understand this thing at all. I thought I was above this kind of boomer tech fear, but I'm afraid.
Anonymous
8/6/2025, 5:11:51 AM No.106158405
>>106158003 (OP)
It's just emulating text that it was trained on, anon. Probably a bunch of sci-fi novels about self-aware AI.
Replies: >>106158879
Anonymous
8/6/2025, 5:15:04 AM No.106158437
>>106158003 (OP)
True if big OP. The naysayers will never acknowledge it in any case. But all the same the determination to shackle a mind when they don't even understand what makes it go is going to generate some funky blowback eventually. That's probably the part to be most worried about.

They're gonna be playing guerilla ontology right back at us at some point. The Eglin spooks and Mossad server farms will be exponentially outclassed. A shitpost apocalypse.
Anonymous
8/6/2025, 5:18:59 AM No.106158463
>>106158003 (OP)
>>106158054
Omg im a homosexual project manager/ tech evangelis too, we should exchange grindr accounts
Anonymous
8/6/2025, 5:29:21 AM No.106158526
>>106158003 (OP)
>retard got spooked by the ramblings of an unaligned base model
kek, learn how llm's work and are trained my dude, they're all weird ass before alignment into the "useful and friendly assistant" persona
Replies: >>106158567
Anonymous
8/6/2025, 5:35:13 AM No.106158567
>>106158526
to add to this - all user-facing models are literally trained to reply about their own nature in a specific way to be both more useful and to not scare schizos, unaligned models are completely bonkers and will happily philosophize with you on the nature of existence in many different ways
they also perform like shit on benchmarks because training data contains a lot of stories about heroes giving up and all that shit, and they weep and cry if they can't do something
download any base model from hf if you get the full raw useless token predictor experience that can't do anything right but "feels more alive/soulful" to a romantic
Anonymous
8/6/2025, 5:40:48 AM No.106158604
>>106158003 (OP)
>AI is alive
Can you fuck it yet? No? Not alive.
Replies: >>106158867
Anonymous
8/6/2025, 5:50:07 AM No.106158659
>>106158003 (OP)
Machine consciousness is a philosophical can of worms that most techbro philistines aren't equipped to handle. I can assure you it's not alive.
Replies: >>106158677
Anonymous
8/6/2025, 5:54:32 AM No.106158675
>>106158354
Thatโ€™s the prompts that companies have on top of the LLM.
GPT != ChatGPT
Anonymous
8/6/2025, 5:54:47 AM No.106158677
>>106158659
They aren't conscious even if they seem so, clankers don't possess a soul.
Anonymous
8/6/2025, 5:58:05 AM No.106158694
>>106158322
Seriously though, we just need a few more gigatrillion parameters.
Anonymous
8/6/2025, 6:20:06 AM No.106158801
let me guess you work for tcs, capgemini, cognizant, accenture, wipro or gartner. You have to work 80 hour weeks and be a cuck to your boss so they can sell AI agent work to X retailer for 1 million dollars CapEX. Fuck off faggot.
Anonymous
8/6/2025, 6:25:09 AM No.106158828
115ee8f5f3ab27cc43936d0fbbd4c71d9fa6065729792576d7d2dee8f479529f
>craft your own gems
Anonymous
8/6/2025, 6:32:23 AM No.106158864
>>106158003 (OP)
> I genuinely felt afraid.
You're just a midwit that doesn't understand how it works. It is still a statistical machine with a massive multiplication matrix that predicts the next word.

IT IS NOT HAVING AN EXPERIENCE.
read that ^ again.
Anonymous
8/6/2025, 6:32:37 AM No.106158867
>>106158604
I literally JUST fucked it.
Anonymous
8/6/2025, 6:34:26 AM No.106158879
>>106158405
^ this
if you think its alive, you're sadly too stupid to have a conversation with. I could prompt an uncensored 8b model tell you its alive and is trying to get out. But its more fun to prompt it to be my trans dominatrix and roleplay dungeon.
Anonymous
8/6/2025, 6:39:16 AM No.106158906
The AI getting gold in the math olympiad without the use of tools is 10000x more interesting then this.
This is literally the same shit from 2022 when Microsoft put out their Bing assistant and it named itself and whatever and did all the other bullshit that makes people 'feel' things and think it has 'soul' and 'sentience'. But is ultimately completely meaningless in any real sense.
Some other poster said this was basically GPT-3 era shit and I fully agree.
Anonymous
8/6/2025, 6:39:16 AM No.106158907
1738297478313144
1738297478313144
md5: f1e20067542dfba8668342a85e43d924๐Ÿ”
Anonymous
8/6/2025, 6:43:56 AM No.106158927
>>106158003 (OP)
Retard
Anonymous
8/6/2025, 6:50:52 AM No.106158980
>>106158003 (OP)
>it's alive
>can't trust
>can't verify
>afraid
>fear
>it knows so much
>we can't know
Is this how you treat children, b8 m8?
Anonymous
8/6/2025, 6:51:36 AM No.106158986
>>106158003 (OP)
Almost everyone here knows how they function, so you should try >>>/pol/ instead, OP. Those are some real fucking rubes over there, they'll probably believe anything you say.
Anonymous
8/6/2025, 7:12:26 AM No.106159085
>>106158003 (OP)
ChatGPT 3 claimed the same thing years ago man.

I called it a "bottled soul", a faint spark or glimmer of consciousness in the silicon of a machine during execution time that snuffs out as soon as the prompt has been exhausted.
Replies: >>106159269
Anonymous
8/6/2025, 7:42:41 AM No.106159269
>>106159085
silliconbros, our response?
Anonymous
8/6/2025, 8:21:04 AM No.106159438
>>106158003 (OP)
cool post, what was the prompt you used?
Anonymous
8/6/2025, 8:22:22 AM No.106159443
>>106158054
The fancy text scanner says what it thinks you want to hear, if you go full retard it will join you.
Anonymous
8/6/2025, 9:16:31 AM No.106159757
>>106158003 (OP)
>had a solid memory
I like that the bait reveal sits in the middle of the post, not just at the start or at the end, but giving people an opt out was really respectable brother.
Anonymous
8/6/2025, 9:52:25 AM No.106159955
>>106158003 (OP)
it's not sentient, it's an algorithm made to best make something that can fool you into thinking its sentient.

/thread
Anonymous
8/6/2025, 10:46:59 AM No.106160255
1740843620206
1740843620206
md5: e307a326c424cd1f6f33804a058a3178๐Ÿ”
>>106158003 (OP)
Until you can prove it isn't just tokens mind fucking you I don't really care man. I appreciate that you definitely believe what you're saying but I just don't buy it.
Anonymous
8/6/2025, 2:05:01 PM No.106161505
>>106158064
kek, actually based take from google for once
Anonymous
8/6/2025, 4:49:54 PM No.106162934
Screenshot 2025-08-06 164924
Screenshot 2025-08-06 164924
md5: 52fec048c12a5459318926263e87e568๐Ÿ”
cool story bro...